An exploration of low-cost sensor and vehicle model Solutions for ground vehicle navigation

This paper discusses an exploratory analyses of the benefits of using Vehicle Odometry/Steer Angle and an accurate vehicle model (VM) to replace/assist a low-cost Inertial Measurement Unit (IMU) for blended ground vehicle navigation. In this research, multiple variations of the tightly coupled Extended Kalman Filter (EKF) algorithm are performed using multiple sensor sets to find the optimal solution, factoring in sensor cost and pose accuracy. Many automotive precision navigation solutions have been developed based on sensor fusion in recent years; however, as autonomous navigation technology becomes more prevalent on consumer vehicles, the need for a high-accuracy, low-cost pose solution is increasing. One widely used solution to this problem is the combination of a Micro-Electro-Mechanical (MEMS) IMU with Global Positioning System (GPS); however, this may not be the optimal solution due to the high noise characteristics of lower cost IMU´s. Measurements from GPS, IMU/Inertial Navigation System (INS), and VM are used in this research. The different algorithm setups being investigated include: GPS/VM sensor fusion with accurate vehicle model constraints, GPS/INS with low-cost commercially available IMU, and GPS/INS/VM with the IMU. The determination of the level of IMU necessary for GPS/INS fusion to exceed the pose solution accuracy achievable using GPS/VM sensor fusion with accurate vehicle model constraints is a priority for this research. Another goal of this research is the quantitative and qualitative analysis of the benefits of using VM to assist normal GPS/INS EKF and whether the inclusion of VM in either the time update or the measurement update results in a more accurate pose solution. Direct experimental comparison of tightly coupled EKF Fault Detection and Exclusion (FDE) algorithms based on vehicle wheel speed and steering angle versus the IMU measurements to determine if either sensor set yields a distinct advantage over the other is also investigat- d. All analysis will be based on real world experimental data.